Zurich's Klarent Lands $8M Seed to Take Deterministic AI Software Testing to the US

The pitch rests on a design choice that Klarent calls deterministic autonomy. Instead of letting an AI agent make live model calls every time a test executes, the platform uses language models only to write and repair the test code. Once written, those tests run the same way each time, and the code is cached and reused, which keeps compute costs down. A human‑in‑the‑loop service reviews each result, and models stay involved whenever a test has to be rewritten after the application changes.

Panakkat has argued that agents which improvise on every run look impressive in a demo but struggle at enterprise scale, where repeatability and cost both matter. One customer reportedly runs around 3,500 tests a day on the platform, which gives a sense of the volume the company is designing for.

The product covers web, iOS, and Android applications. Tests can be generated from plain natural language, adapt when a user interface changes, and plug into CI/CD tools including GitHub, Azure DevOps, Jenkins, and GitLab. Root‑cause findings can be sent to Slack, Microsoft Teams, and Jira. Klarent says it is SOC 2 compliant and can be deployed in a customer's own cloud or on premises, requirements that large buyers tend to check before anything else. It became available on Microsoft Marketplace in September, about a month after the Klarent name was introduced, so the funding lands only weeks after the rebrand. The service is also fully managed, meaning the company handles tests from creation through monitoring and flags regressions to the customer's engineering team, which matters for buyers who would rather not staff a QA function of their own.

Current customers include the retailer JD Sports, the Swiss publisher NZZ, and the car rental company Sixt. At JD Sports, the agents follow a purchase from the moment an order is placed on the website through the order management system and into delivery. Panakkat says in‑house engineers writing tests by hand could not reach that level of coverage. Financial services is the next target, since software failures in that sector carry a high price.

The market argument is straightforward. Panakkat puts quality assurance at roughly a $35 billion business, with most of the work still done by people. Mosaic partner Chandar Lal sees traditional outsourcing, where large teams in lower‑cost locations write test scripts and click through applications, as the real competitor, and described it as a "brute‑force approach" that AI is positioned to displace. He also said the founders recognised the shift while still at Google, and that their commitment to determinism is what persuaded the firm to invest.

On the use of funds, most of the capital will go toward hiring in the US, including a larger sales team across the US and the UK. A smaller share will go to engineering, including cross‑platform testing work.

Klarent is not entering an empty field. Octomind, Momentic, and QA Wolf all sell AI‑driven testing, Tricentis is adding agentic features to an established product, and LambdaTest and Endform are among other companies drawing investor attention. Panakkat contends that rivals built on live model calls will not hold up at enterprise volume, while the established vendors are still retrofitting. That claim will be tested as Klarent meets US buyers who already have a preferred tool.

Pricing has not been spelled out publicly, and the company will need to show that its cost savings against outsourced testing hold up in real contracts. The broader trend is on its side, though. As AI writes a growing share of the world's code, the question of who checks that code becomes harder to ignore. Panakkat expects a future in which one network of agents builds software and another reviews it. Klarent is betting that the reviewing side will matter at least as much as the building side, and investors have now backed that view with $8 million.